We present a deep-learning approach for earthquake detection using waveforms from a seismic array consisting of multiple seismographs. Although automated, deep-learning earthquake detection techniques have recently been developed at the single-station level, they have potential difficulty in reducing false detections owing to the presence of local noise inherent to each station. Here, we propose a deep-learning-based approach to efficiently analyze the waveforms observed by a seismic array, whereby we employ convolutional neural networks in conjunction with graph partitioning to group the waveforms from seismic stations within the array. We then apply the proposed method to waveform data recorded by a dense, local seismic array in the regional seismograph network around the Tokyo metropolitan area, Japan. Our method detects more than $97$ percent of the local seismicity catalogue, with less than $4$ percent false positive rate, based on an optimal threshold value of the output earthquake probability of $0.61$. A comparison with conventional deep-learning-based detectors demonstrates that our method yields fewer false detections for a given true earthquake detection rate. Furthermore, the current method exhibits the robustness to poor-quality data and/or data that are missing at several stations within the array. Synthetic tests demonstrate that the present method has the potential to detect earthquakes even when half of the normally available seismic data are missing. We apply the proposed method to analyze 1-hour-long continuous waveforms and identify new seismic events with extremely low signal-to-noise ratios that are not listed in existing catalogs. (241words)

The discovery of slow slip events (SSEs) based on the installation of dense geodetic observation networks has provided important clues to understanding the process of stress release and accumulation in subduction zones. Because short-term SSEs (S-SSEs) do not often result in sufficient displacements that can be visually inspected, refined automated detection methods are required to understand the occurrence of S-SSEs. In this study, we propose a new method based on which S-SSEs can be detected in observations derived by a Global Navigation Satellite System (GNSS) array by using l1 trend filtering, a variation of sparse estimation, in conjunction with combined -value techniques. The sparse estimation technique and data-driven determination of hyperparameters are utilized in the proposed method to identify candidates of S-SSE onsets. In addition, combined -value techniques are used to provide confidence values for the detections. The results of synthetic tests demonstrated that almost all events can be detected with the new method, with few misdetections, compared with automated detection methods based on Akaike’s information criteria. The proposed method was then applied to daily displacements obtained at 39 GNSS stations in the Nankai subduction zone in western Shikoku, southwest Japan. The results revealed that, in addition to all known events, new events can be detected with the proposed method. Finally, we found the number of low-frequency earthquakes in the target region increased around at the onsets of potential events.